ip6_output.c revision 315514
1/*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $ 30 */ 31 32/*- 33 * Copyright (c) 1982, 1986, 1988, 1990, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 61 */ 62 63#include <sys/cdefs.h> 64__FBSDID("$FreeBSD: stable/11/sys/netinet6/ip6_output.c 315514 2017-03-18 22:04:20Z ae $"); 65 66#include "opt_inet.h" 67#include "opt_inet6.h" 68#include "opt_ipsec.h" 69#include "opt_sctp.h" 70#include "opt_route.h" 71#include "opt_rss.h" 72 73#include <sys/param.h> 74#include <sys/kernel.h> 75#include <sys/malloc.h> 76#include <sys/mbuf.h> 77#include <sys/errno.h> 78#include <sys/priv.h> 79#include <sys/proc.h> 80#include <sys/protosw.h> 81#include <sys/socket.h> 82#include <sys/socketvar.h> 83#include <sys/syslog.h> 84#include <sys/ucred.h> 85 86#include <machine/in_cksum.h> 87 88#include <net/if.h> 89#include <net/if_var.h> 90#include <net/if_llatbl.h> 91#include <net/netisr.h> 92#include <net/route.h> 93#include <net/pfil.h> 94#include <net/rss_config.h> 95#include <net/vnet.h> 96 97#include <netinet/in.h> 98#include <netinet/in_var.h> 99#include <netinet/ip_var.h> 100#include <netinet6/in6_fib.h> 101#include <netinet6/in6_var.h> 102#include <netinet/ip6.h> 103#include <netinet/icmp6.h> 104#include <netinet6/ip6_var.h> 105#include <netinet/in_pcb.h> 106#include <netinet/tcp_var.h> 107#include <netinet6/nd6.h> 108#include <netinet6/in6_rss.h> 109 110#include <netipsec/ipsec_support.h> 111#ifdef SCTP 112#include <netinet/sctp.h> 113#include <netinet/sctp_crc32.h> 114#endif 115 116#include <netinet6/ip6protosw.h> 117#include <netinet6/scope6_var.h> 118 119#ifdef FLOWTABLE 120#include <net/flowtable.h> 121#endif 122 123extern int in6_mcast_loop; 124 125struct ip6_exthdrs { 126 struct mbuf *ip6e_ip6; 127 struct mbuf *ip6e_hbh; 128 struct mbuf *ip6e_dest1; 129 struct mbuf *ip6e_rthdr; 130 struct mbuf *ip6e_dest2; 131}; 132 133static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options"); 134 135static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, 136 struct ucred *, int); 137static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, 138 struct socket *, struct sockopt *); 139static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); 140static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, 141 struct ucred *, int, int, int); 142 143static int ip6_copyexthdr(struct mbuf **, caddr_t, int); 144static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, 145 struct ip6_frag **); 146static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); 147static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); 148static int ip6_getpmtu(struct route_in6 *, int, 149 struct ifnet *, const struct in6_addr *, u_long *, int *, u_int, 150 u_int); 151static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long, 152 u_long *, int *, u_int); 153static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *); 154static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); 155 156 157/* 158 * Make an extension header from option data. hp is the source, and 159 * mp is the destination. 160 */ 161#define MAKE_EXTHDR(hp, mp) \ 162 do { \ 163 if (hp) { \ 164 struct ip6_ext *eh = (struct ip6_ext *)(hp); \ 165 error = ip6_copyexthdr((mp), (caddr_t)(hp), \ 166 ((eh)->ip6e_len + 1) << 3); \ 167 if (error) \ 168 goto freehdrs; \ 169 } \ 170 } while (/*CONSTCOND*/ 0) 171 172/* 173 * Form a chain of extension headers. 174 * m is the extension header mbuf 175 * mp is the previous mbuf in the chain 176 * p is the next header 177 * i is the type of option. 178 */ 179#define MAKE_CHAIN(m, mp, p, i)\ 180 do {\ 181 if (m) {\ 182 if (!hdrsplit) \ 183 panic("assumption failed: hdr not split"); \ 184 *mtod((m), u_char *) = *(p);\ 185 *(p) = (i);\ 186 p = mtod((m), u_char *);\ 187 (m)->m_next = (mp)->m_next;\ 188 (mp)->m_next = (m);\ 189 (mp) = (m);\ 190 }\ 191 } while (/*CONSTCOND*/ 0) 192 193void 194in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset) 195{ 196 u_short csum; 197 198 csum = in_cksum_skip(m, offset + plen, offset); 199 if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0) 200 csum = 0xffff; 201 offset += m->m_pkthdr.csum_data; /* checksum offset */ 202 203 if (offset + sizeof(u_short) > m->m_len) { 204 printf("%s: delayed m_pullup, m->len: %d plen %u off %u " 205 "csum_flags=%b\n", __func__, m->m_len, plen, offset, 206 (int)m->m_pkthdr.csum_flags, CSUM_BITS); 207 /* 208 * XXX this should not happen, but if it does, the correct 209 * behavior may be to insert the checksum in the appropriate 210 * next mbuf in the chain. 211 */ 212 return; 213 } 214 *(u_short *)(m->m_data + offset) = csum; 215} 216 217int 218ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto, 219 int mtu, uint32_t id) 220{ 221 struct mbuf *m, **mnext, *m_frgpart; 222 struct ip6_hdr *ip6, *mhip6; 223 struct ip6_frag *ip6f; 224 int off; 225 int error; 226 int tlen = m0->m_pkthdr.len; 227 228 m = m0; 229 ip6 = mtod(m, struct ip6_hdr *); 230 mnext = &m->m_nextpkt; 231 232 for (off = hlen; off < tlen; off += mtu) { 233 m = m_gethdr(M_NOWAIT, MT_DATA); 234 if (!m) { 235 IP6STAT_INC(ip6s_odropped); 236 return (ENOBUFS); 237 } 238 m->m_flags = m0->m_flags & M_COPYFLAGS; 239 *mnext = m; 240 mnext = &m->m_nextpkt; 241 m->m_data += max_linkhdr; 242 mhip6 = mtod(m, struct ip6_hdr *); 243 *mhip6 = *ip6; 244 m->m_len = sizeof(*mhip6); 245 error = ip6_insertfraghdr(m0, m, hlen, &ip6f); 246 if (error) { 247 IP6STAT_INC(ip6s_odropped); 248 return (error); 249 } 250 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); 251 if (off + mtu >= tlen) 252 mtu = tlen - off; 253 else 254 ip6f->ip6f_offlg |= IP6F_MORE_FRAG; 255 mhip6->ip6_plen = htons((u_short)(mtu + hlen + 256 sizeof(*ip6f) - sizeof(struct ip6_hdr))); 257 if ((m_frgpart = m_copy(m0, off, mtu)) == NULL) { 258 IP6STAT_INC(ip6s_odropped); 259 return (ENOBUFS); 260 } 261 m_cat(m, m_frgpart); 262 m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f); 263 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum; 264 m->m_pkthdr.rcvif = NULL; 265 ip6f->ip6f_reserved = 0; 266 ip6f->ip6f_ident = id; 267 ip6f->ip6f_nxt = nextproto; 268 IP6STAT_INC(ip6s_ofragments); 269 in6_ifstat_inc(ifp, ifs6_out_fragcreat); 270 } 271 272 return (0); 273} 274 275/* 276 * IP6 output. The packet in mbuf chain m contains a skeletal IP6 277 * header (with pri, len, nxt, hlim, src, dst). 278 * This function may modify ver and hlim only. 279 * The mbuf chain containing the packet will be freed. 280 * The mbuf opt, if present, will not be freed. 281 * If route_in6 ro is present and has ro_rt initialized, route lookup would be 282 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, 283 * then result of route lookup is stored in ro->ro_rt. 284 * 285 * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and 286 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one, 287 * which is rt_mtu. 288 * 289 * ifpp - XXX: just for statistics 290 */ 291/* 292 * XXX TODO: no flowid is assigned for outbound flows? 293 */ 294int 295ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, 296 struct route_in6 *ro, int flags, struct ip6_moptions *im6o, 297 struct ifnet **ifpp, struct inpcb *inp) 298{ 299 struct ip6_hdr *ip6; 300 struct ifnet *ifp, *origifp; 301 struct mbuf *m = m0; 302 struct mbuf *mprev = NULL; 303 int hlen, tlen, len; 304 struct route_in6 ip6route; 305 struct rtentry *rt = NULL; 306 struct sockaddr_in6 *dst, src_sa, dst_sa; 307 struct in6_addr odst; 308 int error = 0; 309 struct in6_ifaddr *ia = NULL; 310 u_long mtu; 311 int alwaysfrag, dontfrag; 312 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; 313 struct ip6_exthdrs exthdrs; 314 struct in6_addr src0, dst0; 315 u_int32_t zone; 316 struct route_in6 *ro_pmtu = NULL; 317 int hdrsplit = 0; 318 int sw_csum, tso; 319 int needfiblookup; 320 uint32_t fibnum; 321 struct m_tag *fwd_tag = NULL; 322 uint32_t id; 323 324 if (inp != NULL) { 325 M_SETFIB(m, inp->inp_inc.inc_fibnum); 326 if ((flags & IP_NODEFAULTFLOWID) == 0) { 327 /* unconditionally set flowid */ 328 m->m_pkthdr.flowid = inp->inp_flowid; 329 M_HASHTYPE_SET(m, inp->inp_flowtype); 330 } 331 } 332 333#if defined(IPSEC) || defined(IPSEC_SUPPORT) 334 /* 335 * IPSec checking which handles several cases. 336 * FAST IPSEC: We re-injected the packet. 337 * XXX: need scope argument. 338 */ 339 if (IPSEC_ENABLED(ipv6)) { 340 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) { 341 if (error == EINPROGRESS) 342 error = 0; 343 goto done; 344 } 345 } 346#endif /* IPSEC */ 347 348 bzero(&exthdrs, sizeof(exthdrs)); 349 if (opt) { 350 /* Hop-by-Hop options header */ 351 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); 352 /* Destination options header(1st part) */ 353 if (opt->ip6po_rthdr) { 354 /* 355 * Destination options header(1st part) 356 * This only makes sense with a routing header. 357 * See Section 9.2 of RFC 3542. 358 * Disabling this part just for MIP6 convenience is 359 * a bad idea. We need to think carefully about a 360 * way to make the advanced API coexist with MIP6 361 * options, which might automatically be inserted in 362 * the kernel. 363 */ 364 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); 365 } 366 /* Routing header */ 367 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr); 368 /* Destination options header(2nd part) */ 369 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); 370 } 371 372 /* 373 * Calculate the total length of the extension header chain. 374 * Keep the length of the unfragmentable part for fragmentation. 375 */ 376 optlen = 0; 377 if (exthdrs.ip6e_hbh) 378 optlen += exthdrs.ip6e_hbh->m_len; 379 if (exthdrs.ip6e_dest1) 380 optlen += exthdrs.ip6e_dest1->m_len; 381 if (exthdrs.ip6e_rthdr) 382 optlen += exthdrs.ip6e_rthdr->m_len; 383 unfragpartlen = optlen + sizeof(struct ip6_hdr); 384 385 /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */ 386 if (exthdrs.ip6e_dest2) 387 optlen += exthdrs.ip6e_dest2->m_len; 388 389 /* 390 * If there is at least one extension header, 391 * separate IP6 header from the payload. 392 */ 393 if (optlen && !hdrsplit) { 394 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 395 m = NULL; 396 goto freehdrs; 397 } 398 m = exthdrs.ip6e_ip6; 399 hdrsplit++; 400 } 401 402 ip6 = mtod(m, struct ip6_hdr *); 403 404 /* adjust mbuf packet header length */ 405 m->m_pkthdr.len += optlen; 406 plen = m->m_pkthdr.len - sizeof(*ip6); 407 408 /* If this is a jumbo payload, insert a jumbo payload option. */ 409 if (plen > IPV6_MAXPACKET) { 410 if (!hdrsplit) { 411 if ((error = ip6_splithdr(m, &exthdrs)) != 0) { 412 m = NULL; 413 goto freehdrs; 414 } 415 m = exthdrs.ip6e_ip6; 416 hdrsplit++; 417 } 418 /* adjust pointer */ 419 ip6 = mtod(m, struct ip6_hdr *); 420 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) 421 goto freehdrs; 422 ip6->ip6_plen = 0; 423 } else 424 ip6->ip6_plen = htons(plen); 425 426 /* 427 * Concatenate headers and fill in next header fields. 428 * Here we have, on "m" 429 * IPv6 payload 430 * and we insert headers accordingly. Finally, we should be getting: 431 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload] 432 * 433 * during the header composing process, "m" points to IPv6 header. 434 * "mprev" points to an extension header prior to esp. 435 */ 436 u_char *nexthdrp = &ip6->ip6_nxt; 437 mprev = m; 438 439 /* 440 * we treat dest2 specially. this makes IPsec processing 441 * much easier. the goal here is to make mprev point the 442 * mbuf prior to dest2. 443 * 444 * result: IPv6 dest2 payload 445 * m and mprev will point to IPv6 header. 446 */ 447 if (exthdrs.ip6e_dest2) { 448 if (!hdrsplit) 449 panic("assumption failed: hdr not split"); 450 exthdrs.ip6e_dest2->m_next = m->m_next; 451 m->m_next = exthdrs.ip6e_dest2; 452 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; 453 ip6->ip6_nxt = IPPROTO_DSTOPTS; 454 } 455 456 /* 457 * result: IPv6 hbh dest1 rthdr dest2 payload 458 * m will point to IPv6 header. mprev will point to the 459 * extension header prior to dest2 (rthdr in the above case). 460 */ 461 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); 462 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, 463 IPPROTO_DSTOPTS); 464 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, 465 IPPROTO_ROUTING); 466 467 /* 468 * If there is a routing header, discard the packet. 469 */ 470 if (exthdrs.ip6e_rthdr) { 471 error = EINVAL; 472 goto bad; 473 } 474 475 /* Source address validation */ 476 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && 477 (flags & IPV6_UNSPECSRC) == 0) { 478 error = EOPNOTSUPP; 479 IP6STAT_INC(ip6s_badscope); 480 goto bad; 481 } 482 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 483 error = EOPNOTSUPP; 484 IP6STAT_INC(ip6s_badscope); 485 goto bad; 486 } 487 488 IP6STAT_INC(ip6s_localout); 489 490 /* 491 * Route packet. 492 */ 493 if (ro == NULL) { 494 ro = &ip6route; 495 bzero((caddr_t)ro, sizeof(*ro)); 496 } else 497 ro->ro_flags |= RT_LLE_CACHE; 498 ro_pmtu = ro; 499 if (opt && opt->ip6po_rthdr) 500 ro = &opt->ip6po_route; 501 dst = (struct sockaddr_in6 *)&ro->ro_dst; 502#ifdef FLOWTABLE 503 if (ro->ro_rt == NULL) 504 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro); 505#endif 506 fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m); 507again: 508 /* 509 * if specified, try to fill in the traffic class field. 510 * do not override if a non-zero value is already set. 511 * we check the diffserv field and the ecn field separately. 512 */ 513 if (opt && opt->ip6po_tclass >= 0) { 514 int mask = 0; 515 516 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) 517 mask |= 0xfc; 518 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) 519 mask |= 0x03; 520 if (mask != 0) 521 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); 522 } 523 524 /* fill in or override the hop limit field, if necessary. */ 525 if (opt && opt->ip6po_hlim != -1) 526 ip6->ip6_hlim = opt->ip6po_hlim & 0xff; 527 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 528 if (im6o != NULL) 529 ip6->ip6_hlim = im6o->im6o_multicast_hlim; 530 else 531 ip6->ip6_hlim = V_ip6_defmcasthlim; 532 } 533 /* 534 * Validate route against routing table additions; 535 * a better/more specific route might have been added. 536 * Make sure address family is set in route. 537 */ 538 if (inp) { 539 ro->ro_dst.sin6_family = AF_INET6; 540 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum); 541 } 542 if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) && 543 ro->ro_dst.sin6_family == AF_INET6 && 544 IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) { 545 rt = ro->ro_rt; 546 ifp = ro->ro_rt->rt_ifp; 547 } else { 548 if (ro->ro_lle) 549 LLE_FREE(ro->ro_lle); /* zeros ro_lle */ 550 ro->ro_lle = NULL; 551 if (fwd_tag == NULL) { 552 bzero(&dst_sa, sizeof(dst_sa)); 553 dst_sa.sin6_family = AF_INET6; 554 dst_sa.sin6_len = sizeof(dst_sa); 555 dst_sa.sin6_addr = ip6->ip6_dst; 556 } 557 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp, 558 &rt, fibnum); 559 if (error != 0) { 560 if (ifp != NULL) 561 in6_ifstat_inc(ifp, ifs6_out_discard); 562 goto bad; 563 } 564 } 565 if (rt == NULL) { 566 /* 567 * If in6_selectroute() does not return a route entry, 568 * dst may not have been updated. 569 */ 570 *dst = dst_sa; /* XXX */ 571 } 572 573 /* 574 * then rt (for unicast) and ifp must be non-NULL valid values. 575 */ 576 if ((flags & IPV6_FORWARDING) == 0) { 577 /* XXX: the FORWARDING flag can be set for mrouting. */ 578 in6_ifstat_inc(ifp, ifs6_out_request); 579 } 580 if (rt != NULL) { 581 ia = (struct in6_ifaddr *)(rt->rt_ifa); 582 counter_u64_add(rt->rt_pksent, 1); 583 } 584 585 586 /* 587 * The outgoing interface must be in the zone of source and 588 * destination addresses. 589 */ 590 origifp = ifp; 591 592 src0 = ip6->ip6_src; 593 if (in6_setscope(&src0, origifp, &zone)) 594 goto badscope; 595 bzero(&src_sa, sizeof(src_sa)); 596 src_sa.sin6_family = AF_INET6; 597 src_sa.sin6_len = sizeof(src_sa); 598 src_sa.sin6_addr = ip6->ip6_src; 599 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id) 600 goto badscope; 601 602 dst0 = ip6->ip6_dst; 603 if (in6_setscope(&dst0, origifp, &zone)) 604 goto badscope; 605 /* re-initialize to be sure */ 606 bzero(&dst_sa, sizeof(dst_sa)); 607 dst_sa.sin6_family = AF_INET6; 608 dst_sa.sin6_len = sizeof(dst_sa); 609 dst_sa.sin6_addr = ip6->ip6_dst; 610 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) { 611 goto badscope; 612 } 613 614 /* We should use ia_ifp to support the case of 615 * sending packets to an address of our own. 616 */ 617 if (ia != NULL && ia->ia_ifp) 618 ifp = ia->ia_ifp; 619 620 /* scope check is done. */ 621 goto routefound; 622 623 badscope: 624 IP6STAT_INC(ip6s_badscope); 625 in6_ifstat_inc(origifp, ifs6_out_discard); 626 if (error == 0) 627 error = EHOSTUNREACH; /* XXX */ 628 goto bad; 629 630 routefound: 631 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 632 if (opt && opt->ip6po_nextroute.ro_rt) { 633 /* 634 * The nexthop is explicitly specified by the 635 * application. We assume the next hop is an IPv6 636 * address. 637 */ 638 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop; 639 } 640 else if ((rt->rt_flags & RTF_GATEWAY)) 641 dst = (struct sockaddr_in6 *)rt->rt_gateway; 642 } 643 644 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 645 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ 646 } else { 647 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; 648 in6_ifstat_inc(ifp, ifs6_out_mcast); 649 /* 650 * Confirm that the outgoing interface supports multicast. 651 */ 652 if (!(ifp->if_flags & IFF_MULTICAST)) { 653 IP6STAT_INC(ip6s_noroute); 654 in6_ifstat_inc(ifp, ifs6_out_discard); 655 error = ENETUNREACH; 656 goto bad; 657 } 658 if ((im6o == NULL && in6_mcast_loop) || 659 (im6o && im6o->im6o_multicast_loop)) { 660 /* 661 * Loop back multicast datagram if not expressly 662 * forbidden to do so, even if we have not joined 663 * the address; protocols will filter it later, 664 * thus deferring a hash lookup and lock acquisition 665 * at the expense of an m_copym(). 666 */ 667 ip6_mloopback(ifp, m); 668 } else { 669 /* 670 * If we are acting as a multicast router, perform 671 * multicast forwarding as if the packet had just 672 * arrived on the interface to which we are about 673 * to send. The multicast forwarding function 674 * recursively calls this function, using the 675 * IPV6_FORWARDING flag to prevent infinite recursion. 676 * 677 * Multicasts that are looped back by ip6_mloopback(), 678 * above, will be forwarded by the ip6_input() routine, 679 * if necessary. 680 */ 681 if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { 682 /* 683 * XXX: ip6_mforward expects that rcvif is NULL 684 * when it is called from the originating path. 685 * However, it may not always be the case. 686 */ 687 m->m_pkthdr.rcvif = NULL; 688 if (ip6_mforward(ip6, ifp, m) != 0) { 689 m_freem(m); 690 goto done; 691 } 692 } 693 } 694 /* 695 * Multicasts with a hoplimit of zero may be looped back, 696 * above, but must not be transmitted on a network. 697 * Also, multicasts addressed to the loopback interface 698 * are not sent -- the above call to ip6_mloopback() will 699 * loop back a copy if this host actually belongs to the 700 * destination group on the loopback interface. 701 */ 702 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || 703 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { 704 m_freem(m); 705 goto done; 706 } 707 } 708 709 /* 710 * Fill the outgoing inteface to tell the upper layer 711 * to increment per-interface statistics. 712 */ 713 if (ifpp) 714 *ifpp = ifp; 715 716 /* Determine path MTU. */ 717 if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst, 718 &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0) 719 goto bad; 720 721 /* 722 * The caller of this function may specify to use the minimum MTU 723 * in some cases. 724 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU 725 * setting. The logic is a bit complicated; by default, unicast 726 * packets will follow path MTU while multicast packets will be sent at 727 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets 728 * including unicast ones will be sent at the minimum MTU. Multicast 729 * packets will always be sent at the minimum MTU unless 730 * IP6PO_MINMTU_DISABLE is explicitly specified. 731 * See RFC 3542 for more details. 732 */ 733 if (mtu > IPV6_MMTU) { 734 if ((flags & IPV6_MINMTU)) 735 mtu = IPV6_MMTU; 736 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) 737 mtu = IPV6_MMTU; 738 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 739 (opt == NULL || 740 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { 741 mtu = IPV6_MMTU; 742 } 743 } 744 745 /* 746 * clear embedded scope identifiers if necessary. 747 * in6_clearscope will touch the addresses only when necessary. 748 */ 749 in6_clearscope(&ip6->ip6_src); 750 in6_clearscope(&ip6->ip6_dst); 751 752 /* 753 * If the outgoing packet contains a hop-by-hop options header, 754 * it must be examined and processed even by the source node. 755 * (RFC 2460, section 4.) 756 */ 757 if (exthdrs.ip6e_hbh) { 758 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); 759 u_int32_t dummy; /* XXX unused */ 760 u_int32_t plen = 0; /* XXX: ip6_process will check the value */ 761 762#ifdef DIAGNOSTIC 763 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) 764 panic("ip6e_hbh is not contiguous"); 765#endif 766 /* 767 * XXX: if we have to send an ICMPv6 error to the sender, 768 * we need the M_LOOP flag since icmp6_error() expects 769 * the IPv6 and the hop-by-hop options header are 770 * contiguous unless the flag is set. 771 */ 772 m->m_flags |= M_LOOP; 773 m->m_pkthdr.rcvif = ifp; 774 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), 775 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), 776 &dummy, &plen) < 0) { 777 /* m was already freed at this point */ 778 error = EINVAL;/* better error? */ 779 goto done; 780 } 781 m->m_flags &= ~M_LOOP; /* XXX */ 782 m->m_pkthdr.rcvif = NULL; 783 } 784 785 /* Jump over all PFIL processing if hooks are not active. */ 786 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 787 goto passout; 788 789 odst = ip6->ip6_dst; 790 /* Run through list of hooks for output packets. */ 791 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp); 792 if (error != 0 || m == NULL) 793 goto done; 794 /* adjust pointer */ 795 ip6 = mtod(m, struct ip6_hdr *); 796 797 needfiblookup = 0; 798 /* See if destination IP address was changed by packet filter. */ 799 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { 800 m->m_flags |= M_SKIP_FIREWALL; 801 /* If destination is now ourself drop to ip6_input(). */ 802 if (in6_localip(&ip6->ip6_dst)) { 803 m->m_flags |= M_FASTFWD_OURS; 804 if (m->m_pkthdr.rcvif == NULL) 805 m->m_pkthdr.rcvif = V_loif; 806 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 807 m->m_pkthdr.csum_flags |= 808 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 809 m->m_pkthdr.csum_data = 0xffff; 810 } 811#ifdef SCTP 812 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 813 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 814#endif 815 error = netisr_queue(NETISR_IPV6, m); 816 goto done; 817 } else { 818 RO_RTFREE(ro); 819 needfiblookup = 1; /* Redo the routing table lookup. */ 820 if (ro->ro_lle) 821 LLE_FREE(ro->ro_lle); /* zeros ro_lle */ 822 ro->ro_lle = NULL; 823 } 824 } 825 /* See if fib was changed by packet filter. */ 826 if (fibnum != M_GETFIB(m)) { 827 m->m_flags |= M_SKIP_FIREWALL; 828 fibnum = M_GETFIB(m); 829 RO_RTFREE(ro); 830 needfiblookup = 1; 831 if (ro->ro_lle) 832 LLE_FREE(ro->ro_lle); /* zeros ro_lle */ 833 ro->ro_lle = NULL; 834 } 835 if (needfiblookup) 836 goto again; 837 838 /* See if local, if yes, send it to netisr. */ 839 if (m->m_flags & M_FASTFWD_OURS) { 840 if (m->m_pkthdr.rcvif == NULL) 841 m->m_pkthdr.rcvif = V_loif; 842 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 843 m->m_pkthdr.csum_flags |= 844 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 845 m->m_pkthdr.csum_data = 0xffff; 846 } 847#ifdef SCTP 848 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 849 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 850#endif 851 error = netisr_queue(NETISR_IPV6, m); 852 goto done; 853 } 854 /* Or forward to some other address? */ 855 if ((m->m_flags & M_IP6_NEXTHOP) && 856 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 857 dst = (struct sockaddr_in6 *)&ro->ro_dst; 858 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6)); 859 m->m_flags |= M_SKIP_FIREWALL; 860 m->m_flags &= ~M_IP6_NEXTHOP; 861 m_tag_delete(m, fwd_tag); 862 goto again; 863 } 864 865passout: 866 /* 867 * Send the packet to the outgoing interface. 868 * If necessary, do IPv6 fragmentation before sending. 869 * 870 * the logic here is rather complex: 871 * 1: normal case (dontfrag == 0, alwaysfrag == 0) 872 * 1-a: send as is if tlen <= path mtu 873 * 1-b: fragment if tlen > path mtu 874 * 875 * 2: if user asks us not to fragment (dontfrag == 1) 876 * 2-a: send as is if tlen <= interface mtu 877 * 2-b: error if tlen > interface mtu 878 * 879 * 3: if we always need to attach fragment header (alwaysfrag == 1) 880 * always fragment 881 * 882 * 4: if dontfrag == 1 && alwaysfrag == 1 883 * error, as we cannot handle this conflicting request 884 */ 885 sw_csum = m->m_pkthdr.csum_flags; 886 if (!hdrsplit) { 887 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0; 888 sw_csum &= ~ifp->if_hwassist; 889 } else 890 tso = 0; 891 /* 892 * If we added extension headers, we will not do TSO and calculate the 893 * checksums ourselves for now. 894 * XXX-BZ Need a framework to know when the NIC can handle it, even 895 * with ext. hdrs. 896 */ 897 if (sw_csum & CSUM_DELAY_DATA_IPV6) { 898 sw_csum &= ~CSUM_DELAY_DATA_IPV6; 899 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr)); 900 } 901#ifdef SCTP 902 if (sw_csum & CSUM_SCTP_IPV6) { 903 sw_csum &= ~CSUM_SCTP_IPV6; 904 sctp_delayed_cksum(m, sizeof(struct ip6_hdr)); 905 } 906#endif 907 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 908 tlen = m->m_pkthdr.len; 909 910 if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso) 911 dontfrag = 1; 912 else 913 dontfrag = 0; 914 if (dontfrag && alwaysfrag) { /* case 4 */ 915 /* conflicting request - can't transmit */ 916 error = EMSGSIZE; 917 goto bad; 918 } 919 if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) { /* case 2-b */ 920 /* 921 * Even if the DONTFRAG option is specified, we cannot send the 922 * packet when the data length is larger than the MTU of the 923 * outgoing interface. 924 * Notify the error by sending IPV6_PATHMTU ancillary data if 925 * application wanted to know the MTU value. Also return an 926 * error code (this is not described in the API spec). 927 */ 928 if (inp != NULL) 929 ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu); 930 error = EMSGSIZE; 931 goto bad; 932 } 933 934 /* 935 * transmit packet without fragmentation 936 */ 937 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */ 938 struct in6_ifaddr *ia6; 939 940 ip6 = mtod(m, struct ip6_hdr *); 941 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); 942 if (ia6) { 943 /* Record statistics for this interface address. */ 944 counter_u64_add(ia6->ia_ifa.ifa_opackets, 1); 945 counter_u64_add(ia6->ia_ifa.ifa_obytes, 946 m->m_pkthdr.len); 947 ifa_free(&ia6->ia_ifa); 948 } 949 error = nd6_output_ifp(ifp, origifp, m, dst, 950 (struct route *)ro); 951 goto done; 952 } 953 954 /* 955 * try to fragment the packet. case 1-b and 3 956 */ 957 if (mtu < IPV6_MMTU) { 958 /* path MTU cannot be less than IPV6_MMTU */ 959 error = EMSGSIZE; 960 in6_ifstat_inc(ifp, ifs6_out_fragfail); 961 goto bad; 962 } else if (ip6->ip6_plen == 0) { 963 /* jumbo payload cannot be fragmented */ 964 error = EMSGSIZE; 965 in6_ifstat_inc(ifp, ifs6_out_fragfail); 966 goto bad; 967 } else { 968 u_char nextproto; 969 970 /* 971 * Too large for the destination or interface; 972 * fragment if possible. 973 * Must be able to put at least 8 bytes per fragment. 974 */ 975 hlen = unfragpartlen; 976 if (mtu > IPV6_MAXPACKET) 977 mtu = IPV6_MAXPACKET; 978 979 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; 980 if (len < 8) { 981 error = EMSGSIZE; 982 in6_ifstat_inc(ifp, ifs6_out_fragfail); 983 goto bad; 984 } 985 986 /* 987 * If the interface will not calculate checksums on 988 * fragmented packets, then do it here. 989 * XXX-BZ handle the hw offloading case. Need flags. 990 */ 991 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 992 in6_delayed_cksum(m, plen, hlen); 993 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; 994 } 995#ifdef SCTP 996 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) { 997 sctp_delayed_cksum(m, hlen); 998 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; 999 } 1000#endif 1001 /* 1002 * Change the next header field of the last header in the 1003 * unfragmentable part. 1004 */ 1005 if (exthdrs.ip6e_rthdr) { 1006 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); 1007 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; 1008 } else if (exthdrs.ip6e_dest1) { 1009 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); 1010 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; 1011 } else if (exthdrs.ip6e_hbh) { 1012 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); 1013 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; 1014 } else { 1015 nextproto = ip6->ip6_nxt; 1016 ip6->ip6_nxt = IPPROTO_FRAGMENT; 1017 } 1018 1019 /* 1020 * Loop through length of segment after first fragment, 1021 * make new header and copy data of each part and link onto 1022 * chain. 1023 */ 1024 m0 = m; 1025 id = htonl(ip6_randomid()); 1026 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id))) 1027 goto sendorfree; 1028 1029 in6_ifstat_inc(ifp, ifs6_out_fragok); 1030 } 1031 1032 /* 1033 * Remove leading garbages. 1034 */ 1035sendorfree: 1036 m = m0->m_nextpkt; 1037 m0->m_nextpkt = 0; 1038 m_freem(m0); 1039 for (m0 = m; m; m = m0) { 1040 m0 = m->m_nextpkt; 1041 m->m_nextpkt = 0; 1042 if (error == 0) { 1043 /* Record statistics for this interface address. */ 1044 if (ia) { 1045 counter_u64_add(ia->ia_ifa.ifa_opackets, 1); 1046 counter_u64_add(ia->ia_ifa.ifa_obytes, 1047 m->m_pkthdr.len); 1048 } 1049 error = nd6_output_ifp(ifp, origifp, m, dst, 1050 (struct route *)ro); 1051 } else 1052 m_freem(m); 1053 } 1054 1055 if (error == 0) 1056 IP6STAT_INC(ip6s_fragmented); 1057 1058done: 1059 /* 1060 * Release the route if using our private route, or if 1061 * (with flowtable) we don't have our own reference. 1062 */ 1063 if (ro == &ip6route || 1064 (ro != NULL && ro->ro_flags & RT_NORTREF)) 1065 RO_RTFREE(ro); 1066 return (error); 1067 1068freehdrs: 1069 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ 1070 m_freem(exthdrs.ip6e_dest1); 1071 m_freem(exthdrs.ip6e_rthdr); 1072 m_freem(exthdrs.ip6e_dest2); 1073 /* FALLTHROUGH */ 1074bad: 1075 if (m) 1076 m_freem(m); 1077 goto done; 1078} 1079 1080static int 1081ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) 1082{ 1083 struct mbuf *m; 1084 1085 if (hlen > MCLBYTES) 1086 return (ENOBUFS); /* XXX */ 1087 1088 if (hlen > MLEN) 1089 m = m_getcl(M_NOWAIT, MT_DATA, 0); 1090 else 1091 m = m_get(M_NOWAIT, MT_DATA); 1092 if (m == NULL) 1093 return (ENOBUFS); 1094 m->m_len = hlen; 1095 if (hdr) 1096 bcopy(hdr, mtod(m, caddr_t), hlen); 1097 1098 *mp = m; 1099 return (0); 1100} 1101 1102/* 1103 * Insert jumbo payload option. 1104 */ 1105static int 1106ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 1107{ 1108 struct mbuf *mopt; 1109 u_char *optbuf; 1110 u_int32_t v; 1111 1112#define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ 1113 1114 /* 1115 * If there is no hop-by-hop options header, allocate new one. 1116 * If there is one but it doesn't have enough space to store the 1117 * jumbo payload option, allocate a cluster to store the whole options. 1118 * Otherwise, use it to store the options. 1119 */ 1120 if (exthdrs->ip6e_hbh == NULL) { 1121 mopt = m_get(M_NOWAIT, MT_DATA); 1122 if (mopt == NULL) 1123 return (ENOBUFS); 1124 mopt->m_len = JUMBOOPTLEN; 1125 optbuf = mtod(mopt, u_char *); 1126 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ 1127 exthdrs->ip6e_hbh = mopt; 1128 } else { 1129 struct ip6_hbh *hbh; 1130 1131 mopt = exthdrs->ip6e_hbh; 1132 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { 1133 /* 1134 * XXX assumption: 1135 * - exthdrs->ip6e_hbh is not referenced from places 1136 * other than exthdrs. 1137 * - exthdrs->ip6e_hbh is not an mbuf chain. 1138 */ 1139 int oldoptlen = mopt->m_len; 1140 struct mbuf *n; 1141 1142 /* 1143 * XXX: give up if the whole (new) hbh header does 1144 * not fit even in an mbuf cluster. 1145 */ 1146 if (oldoptlen + JUMBOOPTLEN > MCLBYTES) 1147 return (ENOBUFS); 1148 1149 /* 1150 * As a consequence, we must always prepare a cluster 1151 * at this point. 1152 */ 1153 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1154 if (n == NULL) 1155 return (ENOBUFS); 1156 n->m_len = oldoptlen + JUMBOOPTLEN; 1157 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), 1158 oldoptlen); 1159 optbuf = mtod(n, caddr_t) + oldoptlen; 1160 m_freem(mopt); 1161 mopt = exthdrs->ip6e_hbh = n; 1162 } else { 1163 optbuf = mtod(mopt, u_char *) + mopt->m_len; 1164 mopt->m_len += JUMBOOPTLEN; 1165 } 1166 optbuf[0] = IP6OPT_PADN; 1167 optbuf[1] = 1; 1168 1169 /* 1170 * Adjust the header length according to the pad and 1171 * the jumbo payload option. 1172 */ 1173 hbh = mtod(mopt, struct ip6_hbh *); 1174 hbh->ip6h_len += (JUMBOOPTLEN >> 3); 1175 } 1176 1177 /* fill in the option. */ 1178 optbuf[2] = IP6OPT_JUMBO; 1179 optbuf[3] = 4; 1180 v = (u_int32_t)htonl(plen + JUMBOOPTLEN); 1181 bcopy(&v, &optbuf[4], sizeof(u_int32_t)); 1182 1183 /* finally, adjust the packet header length */ 1184 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; 1185 1186 return (0); 1187#undef JUMBOOPTLEN 1188} 1189 1190/* 1191 * Insert fragment header and copy unfragmentable header portions. 1192 */ 1193static int 1194ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, 1195 struct ip6_frag **frghdrp) 1196{ 1197 struct mbuf *n, *mlast; 1198 1199 if (hlen > sizeof(struct ip6_hdr)) { 1200 n = m_copym(m0, sizeof(struct ip6_hdr), 1201 hlen - sizeof(struct ip6_hdr), M_NOWAIT); 1202 if (n == NULL) 1203 return (ENOBUFS); 1204 m->m_next = n; 1205 } else 1206 n = m; 1207 1208 /* Search for the last mbuf of unfragmentable part. */ 1209 for (mlast = n; mlast->m_next; mlast = mlast->m_next) 1210 ; 1211 1212 if (M_WRITABLE(mlast) && 1213 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) { 1214 /* use the trailing space of the last mbuf for the fragment hdr */ 1215 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + 1216 mlast->m_len); 1217 mlast->m_len += sizeof(struct ip6_frag); 1218 m->m_pkthdr.len += sizeof(struct ip6_frag); 1219 } else { 1220 /* allocate a new mbuf for the fragment header */ 1221 struct mbuf *mfrg; 1222 1223 mfrg = m_get(M_NOWAIT, MT_DATA); 1224 if (mfrg == NULL) 1225 return (ENOBUFS); 1226 mfrg->m_len = sizeof(struct ip6_frag); 1227 *frghdrp = mtod(mfrg, struct ip6_frag *); 1228 mlast->m_next = mfrg; 1229 } 1230 1231 return (0); 1232} 1233 1234/* 1235 * Calculates IPv6 path mtu for destination @dst. 1236 * Resulting MTU is stored in @mtup. 1237 * 1238 * Returns 0 on success. 1239 */ 1240static int 1241ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup) 1242{ 1243 struct nhop6_extended nh6; 1244 struct in6_addr kdst; 1245 uint32_t scopeid; 1246 struct ifnet *ifp; 1247 u_long mtu; 1248 int error; 1249 1250 in6_splitscope(dst, &kdst, &scopeid); 1251 if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0) 1252 return (EHOSTUNREACH); 1253 1254 ifp = nh6.nh_ifp; 1255 mtu = nh6.nh_mtu; 1256 1257 error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0); 1258 fib6_free_nh_ext(fibnum, &nh6); 1259 1260 return (error); 1261} 1262 1263/* 1264 * Calculates IPv6 path MTU for @dst based on transmit @ifp, 1265 * and cached data in @ro_pmtu. 1266 * MTU from (successful) route lookup is saved (along with dst) 1267 * inside @ro_pmtu to avoid subsequent route lookups after packet 1268 * filter processing. 1269 * 1270 * Stores mtu and always-frag value into @mtup and @alwaysfragp. 1271 * Returns 0 on success. 1272 */ 1273static int 1274ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup, 1275 struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup, 1276 int *alwaysfragp, u_int fibnum, u_int proto) 1277{ 1278 struct nhop6_basic nh6; 1279 struct in6_addr kdst; 1280 uint32_t scopeid; 1281 struct sockaddr_in6 *sa6_dst; 1282 u_long mtu; 1283 1284 mtu = 0; 1285 if (do_lookup) { 1286 1287 /* 1288 * Here ro_pmtu has final destination address, while 1289 * ro might represent immediate destination. 1290 * Use ro_pmtu destination since mtu might differ. 1291 */ 1292 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst; 1293 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst)) 1294 ro_pmtu->ro_mtu = 0; 1295 1296 if (ro_pmtu->ro_mtu == 0) { 1297 bzero(sa6_dst, sizeof(*sa6_dst)); 1298 sa6_dst->sin6_family = AF_INET6; 1299 sa6_dst->sin6_len = sizeof(struct sockaddr_in6); 1300 sa6_dst->sin6_addr = *dst; 1301 1302 in6_splitscope(dst, &kdst, &scopeid); 1303 if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0, 1304 &nh6) == 0) 1305 ro_pmtu->ro_mtu = nh6.nh_mtu; 1306 } 1307 1308 mtu = ro_pmtu->ro_mtu; 1309 } 1310 1311 if (ro_pmtu->ro_rt) 1312 mtu = ro_pmtu->ro_rt->rt_mtu; 1313 1314 return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto)); 1315} 1316 1317/* 1318 * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and 1319 * hostcache data for @dst. 1320 * Stores mtu and always-frag value into @mtup and @alwaysfragp. 1321 * 1322 * Returns 0 on success. 1323 */ 1324static int 1325ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu, 1326 u_long *mtup, int *alwaysfragp, u_int proto) 1327{ 1328 u_long mtu = 0; 1329 int alwaysfrag = 0; 1330 int error = 0; 1331 1332 if (rt_mtu > 0) { 1333 u_int32_t ifmtu; 1334 struct in_conninfo inc; 1335 1336 bzero(&inc, sizeof(inc)); 1337 inc.inc_flags |= INC_ISIPV6; 1338 inc.inc6_faddr = *dst; 1339 1340 ifmtu = IN6_LINKMTU(ifp); 1341 1342 /* TCP is known to react to pmtu changes so skip hc */ 1343 if (proto != IPPROTO_TCP) 1344 mtu = tcp_hc_getmtu(&inc); 1345 1346 if (mtu) 1347 mtu = min(mtu, rt_mtu); 1348 else 1349 mtu = rt_mtu; 1350 if (mtu == 0) 1351 mtu = ifmtu; 1352 else if (mtu < IPV6_MMTU) { 1353 /* 1354 * RFC2460 section 5, last paragraph: 1355 * if we record ICMPv6 too big message with 1356 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU 1357 * or smaller, with framgent header attached. 1358 * (fragment header is needed regardless from the 1359 * packet size, for translators to identify packets) 1360 */ 1361 alwaysfrag = 1; 1362 mtu = IPV6_MMTU; 1363 } 1364 } else if (ifp) { 1365 mtu = IN6_LINKMTU(ifp); 1366 } else 1367 error = EHOSTUNREACH; /* XXX */ 1368 1369 *mtup = mtu; 1370 if (alwaysfragp) 1371 *alwaysfragp = alwaysfrag; 1372 return (error); 1373} 1374 1375/* 1376 * IP6 socket option processing. 1377 */ 1378int 1379ip6_ctloutput(struct socket *so, struct sockopt *sopt) 1380{ 1381 int optdatalen, uproto; 1382 void *optdata; 1383 struct inpcb *in6p = sotoinpcb(so); 1384 int error, optval; 1385 int level, op, optname; 1386 int optlen; 1387 struct thread *td; 1388#ifdef RSS 1389 uint32_t rss_bucket; 1390 int retval; 1391#endif 1392 1393/* 1394 * Don't use more than a quarter of mbuf clusters. N.B.: 1395 * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow 1396 * on LP64 architectures, so cast to u_long to avoid undefined 1397 * behavior. ILP32 architectures cannot have nmbclusters 1398 * large enough to overflow for other reasons. 1399 */ 1400#define IPV6_PKTOPTIONS_MBUF_LIMIT ((u_long)nmbclusters * MCLBYTES / 4) 1401 1402 level = sopt->sopt_level; 1403 op = sopt->sopt_dir; 1404 optname = sopt->sopt_name; 1405 optlen = sopt->sopt_valsize; 1406 td = sopt->sopt_td; 1407 error = 0; 1408 optval = 0; 1409 uproto = (int)so->so_proto->pr_protocol; 1410 1411 if (level != IPPROTO_IPV6) { 1412 error = EINVAL; 1413 1414 if (sopt->sopt_level == SOL_SOCKET && 1415 sopt->sopt_dir == SOPT_SET) { 1416 switch (sopt->sopt_name) { 1417 case SO_REUSEADDR: 1418 INP_WLOCK(in6p); 1419 if ((so->so_options & SO_REUSEADDR) != 0) 1420 in6p->inp_flags2 |= INP_REUSEADDR; 1421 else 1422 in6p->inp_flags2 &= ~INP_REUSEADDR; 1423 INP_WUNLOCK(in6p); 1424 error = 0; 1425 break; 1426 case SO_REUSEPORT: 1427 INP_WLOCK(in6p); 1428 if ((so->so_options & SO_REUSEPORT) != 0) 1429 in6p->inp_flags2 |= INP_REUSEPORT; 1430 else 1431 in6p->inp_flags2 &= ~INP_REUSEPORT; 1432 INP_WUNLOCK(in6p); 1433 error = 0; 1434 break; 1435 case SO_SETFIB: 1436 INP_WLOCK(in6p); 1437 in6p->inp_inc.inc_fibnum = so->so_fibnum; 1438 INP_WUNLOCK(in6p); 1439 error = 0; 1440 break; 1441 default: 1442 break; 1443 } 1444 } 1445 } else { /* level == IPPROTO_IPV6 */ 1446 switch (op) { 1447 1448 case SOPT_SET: 1449 switch (optname) { 1450 case IPV6_2292PKTOPTIONS: 1451#ifdef IPV6_PKTOPTIONS 1452 case IPV6_PKTOPTIONS: 1453#endif 1454 { 1455 struct mbuf *m; 1456 1457 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) { 1458 printf("ip6_ctloutput: mbuf limit hit\n"); 1459 error = ENOBUFS; 1460 break; 1461 } 1462 1463 error = soopt_getm(sopt, &m); /* XXX */ 1464 if (error != 0) 1465 break; 1466 error = soopt_mcopyin(sopt, m); /* XXX */ 1467 if (error != 0) 1468 break; 1469 error = ip6_pcbopts(&in6p->in6p_outputopts, 1470 m, so, sopt); 1471 m_freem(m); /* XXX */ 1472 break; 1473 } 1474 1475 /* 1476 * Use of some Hop-by-Hop options or some 1477 * Destination options, might require special 1478 * privilege. That is, normal applications 1479 * (without special privilege) might be forbidden 1480 * from setting certain options in outgoing packets, 1481 * and might never see certain options in received 1482 * packets. [RFC 2292 Section 6] 1483 * KAME specific note: 1484 * KAME prevents non-privileged users from sending or 1485 * receiving ANY hbh/dst options in order to avoid 1486 * overhead of parsing options in the kernel. 1487 */ 1488 case IPV6_RECVHOPOPTS: 1489 case IPV6_RECVDSTOPTS: 1490 case IPV6_RECVRTHDRDSTOPTS: 1491 if (td != NULL) { 1492 error = priv_check(td, 1493 PRIV_NETINET_SETHDROPTS); 1494 if (error) 1495 break; 1496 } 1497 /* FALLTHROUGH */ 1498 case IPV6_UNICAST_HOPS: 1499 case IPV6_HOPLIMIT: 1500 1501 case IPV6_RECVPKTINFO: 1502 case IPV6_RECVHOPLIMIT: 1503 case IPV6_RECVRTHDR: 1504 case IPV6_RECVPATHMTU: 1505 case IPV6_RECVTCLASS: 1506 case IPV6_RECVFLOWID: 1507#ifdef RSS 1508 case IPV6_RECVRSSBUCKETID: 1509#endif 1510 case IPV6_V6ONLY: 1511 case IPV6_AUTOFLOWLABEL: 1512 case IPV6_BINDANY: 1513 case IPV6_BINDMULTI: 1514#ifdef RSS 1515 case IPV6_RSS_LISTEN_BUCKET: 1516#endif 1517 if (optname == IPV6_BINDANY && td != NULL) { 1518 error = priv_check(td, 1519 PRIV_NETINET_BINDANY); 1520 if (error) 1521 break; 1522 } 1523 1524 if (optlen != sizeof(int)) { 1525 error = EINVAL; 1526 break; 1527 } 1528 error = sooptcopyin(sopt, &optval, 1529 sizeof optval, sizeof optval); 1530 if (error) 1531 break; 1532 switch (optname) { 1533 1534 case IPV6_UNICAST_HOPS: 1535 if (optval < -1 || optval >= 256) 1536 error = EINVAL; 1537 else { 1538 /* -1 = kernel default */ 1539 in6p->in6p_hops = optval; 1540 if ((in6p->inp_vflag & 1541 INP_IPV4) != 0) 1542 in6p->inp_ip_ttl = optval; 1543 } 1544 break; 1545#define OPTSET(bit) \ 1546do { \ 1547 INP_WLOCK(in6p); \ 1548 if (optval) \ 1549 in6p->inp_flags |= (bit); \ 1550 else \ 1551 in6p->inp_flags &= ~(bit); \ 1552 INP_WUNLOCK(in6p); \ 1553} while (/*CONSTCOND*/ 0) 1554#define OPTSET2292(bit) \ 1555do { \ 1556 INP_WLOCK(in6p); \ 1557 in6p->inp_flags |= IN6P_RFC2292; \ 1558 if (optval) \ 1559 in6p->inp_flags |= (bit); \ 1560 else \ 1561 in6p->inp_flags &= ~(bit); \ 1562 INP_WUNLOCK(in6p); \ 1563} while (/*CONSTCOND*/ 0) 1564#define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0) 1565 1566#define OPTSET2(bit, val) do { \ 1567 INP_WLOCK(in6p); \ 1568 if (val) \ 1569 in6p->inp_flags2 |= bit; \ 1570 else \ 1571 in6p->inp_flags2 &= ~bit; \ 1572 INP_WUNLOCK(in6p); \ 1573} while (0) 1574#define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0) 1575 1576 case IPV6_RECVPKTINFO: 1577 /* cannot mix with RFC2292 */ 1578 if (OPTBIT(IN6P_RFC2292)) { 1579 error = EINVAL; 1580 break; 1581 } 1582 OPTSET(IN6P_PKTINFO); 1583 break; 1584 1585 case IPV6_HOPLIMIT: 1586 { 1587 struct ip6_pktopts **optp; 1588 1589 /* cannot mix with RFC2292 */ 1590 if (OPTBIT(IN6P_RFC2292)) { 1591 error = EINVAL; 1592 break; 1593 } 1594 optp = &in6p->in6p_outputopts; 1595 error = ip6_pcbopt(IPV6_HOPLIMIT, 1596 (u_char *)&optval, sizeof(optval), 1597 optp, (td != NULL) ? td->td_ucred : 1598 NULL, uproto); 1599 break; 1600 } 1601 1602 case IPV6_RECVHOPLIMIT: 1603 /* cannot mix with RFC2292 */ 1604 if (OPTBIT(IN6P_RFC2292)) { 1605 error = EINVAL; 1606 break; 1607 } 1608 OPTSET(IN6P_HOPLIMIT); 1609 break; 1610 1611 case IPV6_RECVHOPOPTS: 1612 /* cannot mix with RFC2292 */ 1613 if (OPTBIT(IN6P_RFC2292)) { 1614 error = EINVAL; 1615 break; 1616 } 1617 OPTSET(IN6P_HOPOPTS); 1618 break; 1619 1620 case IPV6_RECVDSTOPTS: 1621 /* cannot mix with RFC2292 */ 1622 if (OPTBIT(IN6P_RFC2292)) { 1623 error = EINVAL; 1624 break; 1625 } 1626 OPTSET(IN6P_DSTOPTS); 1627 break; 1628 1629 case IPV6_RECVRTHDRDSTOPTS: 1630 /* cannot mix with RFC2292 */ 1631 if (OPTBIT(IN6P_RFC2292)) { 1632 error = EINVAL; 1633 break; 1634 } 1635 OPTSET(IN6P_RTHDRDSTOPTS); 1636 break; 1637 1638 case IPV6_RECVRTHDR: 1639 /* cannot mix with RFC2292 */ 1640 if (OPTBIT(IN6P_RFC2292)) { 1641 error = EINVAL; 1642 break; 1643 } 1644 OPTSET(IN6P_RTHDR); 1645 break; 1646 1647 case IPV6_RECVPATHMTU: 1648 /* 1649 * We ignore this option for TCP 1650 * sockets. 1651 * (RFC3542 leaves this case 1652 * unspecified.) 1653 */ 1654 if (uproto != IPPROTO_TCP) 1655 OPTSET(IN6P_MTU); 1656 break; 1657 1658 case IPV6_RECVFLOWID: 1659 OPTSET2(INP_RECVFLOWID, optval); 1660 break; 1661 1662#ifdef RSS 1663 case IPV6_RECVRSSBUCKETID: 1664 OPTSET2(INP_RECVRSSBUCKETID, optval); 1665 break; 1666#endif 1667 1668 case IPV6_V6ONLY: 1669 /* 1670 * make setsockopt(IPV6_V6ONLY) 1671 * available only prior to bind(2). 1672 * see ipng mailing list, Jun 22 2001. 1673 */ 1674 if (in6p->inp_lport || 1675 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { 1676 error = EINVAL; 1677 break; 1678 } 1679 OPTSET(IN6P_IPV6_V6ONLY); 1680 if (optval) 1681 in6p->inp_vflag &= ~INP_IPV4; 1682 else 1683 in6p->inp_vflag |= INP_IPV4; 1684 break; 1685 case IPV6_RECVTCLASS: 1686 /* cannot mix with RFC2292 XXX */ 1687 if (OPTBIT(IN6P_RFC2292)) { 1688 error = EINVAL; 1689 break; 1690 } 1691 OPTSET(IN6P_TCLASS); 1692 break; 1693 case IPV6_AUTOFLOWLABEL: 1694 OPTSET(IN6P_AUTOFLOWLABEL); 1695 break; 1696 1697 case IPV6_BINDANY: 1698 OPTSET(INP_BINDANY); 1699 break; 1700 1701 case IPV6_BINDMULTI: 1702 OPTSET2(INP_BINDMULTI, optval); 1703 break; 1704#ifdef RSS 1705 case IPV6_RSS_LISTEN_BUCKET: 1706 if ((optval >= 0) && 1707 (optval < rss_getnumbuckets())) { 1708 in6p->inp_rss_listen_bucket = optval; 1709 OPTSET2(INP_RSS_BUCKET_SET, 1); 1710 } else { 1711 error = EINVAL; 1712 } 1713 break; 1714#endif 1715 } 1716 break; 1717 1718 case IPV6_TCLASS: 1719 case IPV6_DONTFRAG: 1720 case IPV6_USE_MIN_MTU: 1721 case IPV6_PREFER_TEMPADDR: 1722 if (optlen != sizeof(optval)) { 1723 error = EINVAL; 1724 break; 1725 } 1726 error = sooptcopyin(sopt, &optval, 1727 sizeof optval, sizeof optval); 1728 if (error) 1729 break; 1730 { 1731 struct ip6_pktopts **optp; 1732 optp = &in6p->in6p_outputopts; 1733 error = ip6_pcbopt(optname, 1734 (u_char *)&optval, sizeof(optval), 1735 optp, (td != NULL) ? td->td_ucred : 1736 NULL, uproto); 1737 break; 1738 } 1739 1740 case IPV6_2292PKTINFO: 1741 case IPV6_2292HOPLIMIT: 1742 case IPV6_2292HOPOPTS: 1743 case IPV6_2292DSTOPTS: 1744 case IPV6_2292RTHDR: 1745 /* RFC 2292 */ 1746 if (optlen != sizeof(int)) { 1747 error = EINVAL; 1748 break; 1749 } 1750 error = sooptcopyin(sopt, &optval, 1751 sizeof optval, sizeof optval); 1752 if (error) 1753 break; 1754 switch (optname) { 1755 case IPV6_2292PKTINFO: 1756 OPTSET2292(IN6P_PKTINFO); 1757 break; 1758 case IPV6_2292HOPLIMIT: 1759 OPTSET2292(IN6P_HOPLIMIT); 1760 break; 1761 case IPV6_2292HOPOPTS: 1762 /* 1763 * Check super-user privilege. 1764 * See comments for IPV6_RECVHOPOPTS. 1765 */ 1766 if (td != NULL) { 1767 error = priv_check(td, 1768 PRIV_NETINET_SETHDROPTS); 1769 if (error) 1770 return (error); 1771 } 1772 OPTSET2292(IN6P_HOPOPTS); 1773 break; 1774 case IPV6_2292DSTOPTS: 1775 if (td != NULL) { 1776 error = priv_check(td, 1777 PRIV_NETINET_SETHDROPTS); 1778 if (error) 1779 return (error); 1780 } 1781 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ 1782 break; 1783 case IPV6_2292RTHDR: 1784 OPTSET2292(IN6P_RTHDR); 1785 break; 1786 } 1787 break; 1788 case IPV6_PKTINFO: 1789 case IPV6_HOPOPTS: 1790 case IPV6_RTHDR: 1791 case IPV6_DSTOPTS: 1792 case IPV6_RTHDRDSTOPTS: 1793 case IPV6_NEXTHOP: 1794 { 1795 /* new advanced API (RFC3542) */ 1796 u_char *optbuf; 1797 u_char optbuf_storage[MCLBYTES]; 1798 int optlen; 1799 struct ip6_pktopts **optp; 1800 1801 /* cannot mix with RFC2292 */ 1802 if (OPTBIT(IN6P_RFC2292)) { 1803 error = EINVAL; 1804 break; 1805 } 1806 1807 /* 1808 * We only ensure valsize is not too large 1809 * here. Further validation will be done 1810 * later. 1811 */ 1812 error = sooptcopyin(sopt, optbuf_storage, 1813 sizeof(optbuf_storage), 0); 1814 if (error) 1815 break; 1816 optlen = sopt->sopt_valsize; 1817 optbuf = optbuf_storage; 1818 optp = &in6p->in6p_outputopts; 1819 error = ip6_pcbopt(optname, optbuf, optlen, 1820 optp, (td != NULL) ? td->td_ucred : NULL, 1821 uproto); 1822 break; 1823 } 1824#undef OPTSET 1825 1826 case IPV6_MULTICAST_IF: 1827 case IPV6_MULTICAST_HOPS: 1828 case IPV6_MULTICAST_LOOP: 1829 case IPV6_JOIN_GROUP: 1830 case IPV6_LEAVE_GROUP: 1831 case IPV6_MSFILTER: 1832 case MCAST_BLOCK_SOURCE: 1833 case MCAST_UNBLOCK_SOURCE: 1834 case MCAST_JOIN_GROUP: 1835 case MCAST_LEAVE_GROUP: 1836 case MCAST_JOIN_SOURCE_GROUP: 1837 case MCAST_LEAVE_SOURCE_GROUP: 1838 error = ip6_setmoptions(in6p, sopt); 1839 break; 1840 1841 case IPV6_PORTRANGE: 1842 error = sooptcopyin(sopt, &optval, 1843 sizeof optval, sizeof optval); 1844 if (error) 1845 break; 1846 1847 INP_WLOCK(in6p); 1848 switch (optval) { 1849 case IPV6_PORTRANGE_DEFAULT: 1850 in6p->inp_flags &= ~(INP_LOWPORT); 1851 in6p->inp_flags &= ~(INP_HIGHPORT); 1852 break; 1853 1854 case IPV6_PORTRANGE_HIGH: 1855 in6p->inp_flags &= ~(INP_LOWPORT); 1856 in6p->inp_flags |= INP_HIGHPORT; 1857 break; 1858 1859 case IPV6_PORTRANGE_LOW: 1860 in6p->inp_flags &= ~(INP_HIGHPORT); 1861 in6p->inp_flags |= INP_LOWPORT; 1862 break; 1863 1864 default: 1865 error = EINVAL; 1866 break; 1867 } 1868 INP_WUNLOCK(in6p); 1869 break; 1870 1871#if defined(IPSEC) || defined(IPSEC_SUPPORT) 1872 case IPV6_IPSEC_POLICY: 1873 if (IPSEC_ENABLED(ipv6)) { 1874 error = IPSEC_PCBCTL(ipv6, in6p, sopt); 1875 break; 1876 } 1877 /* FALLTHROUGH */ 1878#endif /* IPSEC */ 1879 1880 default: 1881 error = ENOPROTOOPT; 1882 break; 1883 } 1884 break; 1885 1886 case SOPT_GET: 1887 switch (optname) { 1888 1889 case IPV6_2292PKTOPTIONS: 1890#ifdef IPV6_PKTOPTIONS 1891 case IPV6_PKTOPTIONS: 1892#endif 1893 /* 1894 * RFC3542 (effectively) deprecated the 1895 * semantics of the 2292-style pktoptions. 1896 * Since it was not reliable in nature (i.e., 1897 * applications had to expect the lack of some 1898 * information after all), it would make sense 1899 * to simplify this part by always returning 1900 * empty data. 1901 */ 1902 sopt->sopt_valsize = 0; 1903 break; 1904 1905 case IPV6_RECVHOPOPTS: 1906 case IPV6_RECVDSTOPTS: 1907 case IPV6_RECVRTHDRDSTOPTS: 1908 case IPV6_UNICAST_HOPS: 1909 case IPV6_RECVPKTINFO: 1910 case IPV6_RECVHOPLIMIT: 1911 case IPV6_RECVRTHDR: 1912 case IPV6_RECVPATHMTU: 1913 1914 case IPV6_V6ONLY: 1915 case IPV6_PORTRANGE: 1916 case IPV6_RECVTCLASS: 1917 case IPV6_AUTOFLOWLABEL: 1918 case IPV6_BINDANY: 1919 case IPV6_FLOWID: 1920 case IPV6_FLOWTYPE: 1921 case IPV6_RECVFLOWID: 1922#ifdef RSS 1923 case IPV6_RSSBUCKETID: 1924 case IPV6_RECVRSSBUCKETID: 1925#endif 1926 case IPV6_BINDMULTI: 1927 switch (optname) { 1928 1929 case IPV6_RECVHOPOPTS: 1930 optval = OPTBIT(IN6P_HOPOPTS); 1931 break; 1932 1933 case IPV6_RECVDSTOPTS: 1934 optval = OPTBIT(IN6P_DSTOPTS); 1935 break; 1936 1937 case IPV6_RECVRTHDRDSTOPTS: 1938 optval = OPTBIT(IN6P_RTHDRDSTOPTS); 1939 break; 1940 1941 case IPV6_UNICAST_HOPS: 1942 optval = in6p->in6p_hops; 1943 break; 1944 1945 case IPV6_RECVPKTINFO: 1946 optval = OPTBIT(IN6P_PKTINFO); 1947 break; 1948 1949 case IPV6_RECVHOPLIMIT: 1950 optval = OPTBIT(IN6P_HOPLIMIT); 1951 break; 1952 1953 case IPV6_RECVRTHDR: 1954 optval = OPTBIT(IN6P_RTHDR); 1955 break; 1956 1957 case IPV6_RECVPATHMTU: 1958 optval = OPTBIT(IN6P_MTU); 1959 break; 1960 1961 case IPV6_V6ONLY: 1962 optval = OPTBIT(IN6P_IPV6_V6ONLY); 1963 break; 1964 1965 case IPV6_PORTRANGE: 1966 { 1967 int flags; 1968 flags = in6p->inp_flags; 1969 if (flags & INP_HIGHPORT) 1970 optval = IPV6_PORTRANGE_HIGH; 1971 else if (flags & INP_LOWPORT) 1972 optval = IPV6_PORTRANGE_LOW; 1973 else 1974 optval = 0; 1975 break; 1976 } 1977 case IPV6_RECVTCLASS: 1978 optval = OPTBIT(IN6P_TCLASS); 1979 break; 1980 1981 case IPV6_AUTOFLOWLABEL: 1982 optval = OPTBIT(IN6P_AUTOFLOWLABEL); 1983 break; 1984 1985 case IPV6_BINDANY: 1986 optval = OPTBIT(INP_BINDANY); 1987 break; 1988 1989 case IPV6_FLOWID: 1990 optval = in6p->inp_flowid; 1991 break; 1992 1993 case IPV6_FLOWTYPE: 1994 optval = in6p->inp_flowtype; 1995 break; 1996 1997 case IPV6_RECVFLOWID: 1998 optval = OPTBIT2(INP_RECVFLOWID); 1999 break; 2000#ifdef RSS 2001 case IPV6_RSSBUCKETID: 2002 retval = 2003 rss_hash2bucket(in6p->inp_flowid, 2004 in6p->inp_flowtype, 2005 &rss_bucket); 2006 if (retval == 0) 2007 optval = rss_bucket; 2008 else 2009 error = EINVAL; 2010 break; 2011 2012 case IPV6_RECVRSSBUCKETID: 2013 optval = OPTBIT2(INP_RECVRSSBUCKETID); 2014 break; 2015#endif 2016 2017 case IPV6_BINDMULTI: 2018 optval = OPTBIT2(INP_BINDMULTI); 2019 break; 2020 2021 } 2022 if (error) 2023 break; 2024 error = sooptcopyout(sopt, &optval, 2025 sizeof optval); 2026 break; 2027 2028 case IPV6_PATHMTU: 2029 { 2030 u_long pmtu = 0; 2031 struct ip6_mtuinfo mtuinfo; 2032 2033 if (!(so->so_state & SS_ISCONNECTED)) 2034 return (ENOTCONN); 2035 /* 2036 * XXX: we dot not consider the case of source 2037 * routing, or optional information to specify 2038 * the outgoing interface. 2039 */ 2040 error = ip6_getpmtu_ctl(so->so_fibnum, 2041 &in6p->in6p_faddr, &pmtu); 2042 if (error) 2043 break; 2044 if (pmtu > IPV6_MAXPACKET) 2045 pmtu = IPV6_MAXPACKET; 2046 2047 bzero(&mtuinfo, sizeof(mtuinfo)); 2048 mtuinfo.ip6m_mtu = (u_int32_t)pmtu; 2049 optdata = (void *)&mtuinfo; 2050 optdatalen = sizeof(mtuinfo); 2051 error = sooptcopyout(sopt, optdata, 2052 optdatalen); 2053 break; 2054 } 2055 2056 case IPV6_2292PKTINFO: 2057 case IPV6_2292HOPLIMIT: 2058 case IPV6_2292HOPOPTS: 2059 case IPV6_2292RTHDR: 2060 case IPV6_2292DSTOPTS: 2061 switch (optname) { 2062 case IPV6_2292PKTINFO: 2063 optval = OPTBIT(IN6P_PKTINFO); 2064 break; 2065 case IPV6_2292HOPLIMIT: 2066 optval = OPTBIT(IN6P_HOPLIMIT); 2067 break; 2068 case IPV6_2292HOPOPTS: 2069 optval = OPTBIT(IN6P_HOPOPTS); 2070 break; 2071 case IPV6_2292RTHDR: 2072 optval = OPTBIT(IN6P_RTHDR); 2073 break; 2074 case IPV6_2292DSTOPTS: 2075 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); 2076 break; 2077 } 2078 error = sooptcopyout(sopt, &optval, 2079 sizeof optval); 2080 break; 2081 case IPV6_PKTINFO: 2082 case IPV6_HOPOPTS: 2083 case IPV6_RTHDR: 2084 case IPV6_DSTOPTS: 2085 case IPV6_RTHDRDSTOPTS: 2086 case IPV6_NEXTHOP: 2087 case IPV6_TCLASS: 2088 case IPV6_DONTFRAG: 2089 case IPV6_USE_MIN_MTU: 2090 case IPV6_PREFER_TEMPADDR: 2091 error = ip6_getpcbopt(in6p->in6p_outputopts, 2092 optname, sopt); 2093 break; 2094 2095 case IPV6_MULTICAST_IF: 2096 case IPV6_MULTICAST_HOPS: 2097 case IPV6_MULTICAST_LOOP: 2098 case IPV6_MSFILTER: 2099 error = ip6_getmoptions(in6p, sopt); 2100 break; 2101 2102#if defined(IPSEC) || defined(IPSEC_SUPPORT) 2103 case IPV6_IPSEC_POLICY: 2104 if (IPSEC_ENABLED(ipv6)) { 2105 error = IPSEC_PCBCTL(ipv6, in6p, sopt); 2106 break; 2107 } 2108 /* FALLTHROUGH */ 2109#endif /* IPSEC */ 2110 default: 2111 error = ENOPROTOOPT; 2112 break; 2113 } 2114 break; 2115 } 2116 } 2117 return (error); 2118} 2119 2120int 2121ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) 2122{ 2123 int error = 0, optval, optlen; 2124 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); 2125 struct inpcb *in6p = sotoinpcb(so); 2126 int level, op, optname; 2127 2128 level = sopt->sopt_level; 2129 op = sopt->sopt_dir; 2130 optname = sopt->sopt_name; 2131 optlen = sopt->sopt_valsize; 2132 2133 if (level != IPPROTO_IPV6) { 2134 return (EINVAL); 2135 } 2136 2137 switch (optname) { 2138 case IPV6_CHECKSUM: 2139 /* 2140 * For ICMPv6 sockets, no modification allowed for checksum 2141 * offset, permit "no change" values to help existing apps. 2142 * 2143 * RFC3542 says: "An attempt to set IPV6_CHECKSUM 2144 * for an ICMPv6 socket will fail." 2145 * The current behavior does not meet RFC3542. 2146 */ 2147 switch (op) { 2148 case SOPT_SET: 2149 if (optlen != sizeof(int)) { 2150 error = EINVAL; 2151 break; 2152 } 2153 error = sooptcopyin(sopt, &optval, sizeof(optval), 2154 sizeof(optval)); 2155 if (error) 2156 break; 2157 if ((optval % 2) != 0) { 2158 /* the API assumes even offset values */ 2159 error = EINVAL; 2160 } else if (so->so_proto->pr_protocol == 2161 IPPROTO_ICMPV6) { 2162 if (optval != icmp6off) 2163 error = EINVAL; 2164 } else 2165 in6p->in6p_cksum = optval; 2166 break; 2167 2168 case SOPT_GET: 2169 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) 2170 optval = icmp6off; 2171 else 2172 optval = in6p->in6p_cksum; 2173 2174 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2175 break; 2176 2177 default: 2178 error = EINVAL; 2179 break; 2180 } 2181 break; 2182 2183 default: 2184 error = ENOPROTOOPT; 2185 break; 2186 } 2187 2188 return (error); 2189} 2190 2191/* 2192 * Set up IP6 options in pcb for insertion in output packets or 2193 * specifying behavior of outgoing packets. 2194 */ 2195static int 2196ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, 2197 struct socket *so, struct sockopt *sopt) 2198{ 2199 struct ip6_pktopts *opt = *pktopt; 2200 int error = 0; 2201 struct thread *td = sopt->sopt_td; 2202 2203 /* turn off any old options. */ 2204 if (opt) { 2205#ifdef DIAGNOSTIC 2206 if (opt->ip6po_pktinfo || opt->ip6po_nexthop || 2207 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || 2208 opt->ip6po_rhinfo.ip6po_rhi_rthdr) 2209 printf("ip6_pcbopts: all specified options are cleared.\n"); 2210#endif 2211 ip6_clearpktopts(opt, -1); 2212 } else 2213 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK); 2214 *pktopt = NULL; 2215 2216 if (!m || m->m_len == 0) { 2217 /* 2218 * Only turning off any previous options, regardless of 2219 * whether the opt is just created or given. 2220 */ 2221 free(opt, M_IP6OPT); 2222 return (0); 2223 } 2224 2225 /* set options specified by user. */ 2226 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ? 2227 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) { 2228 ip6_clearpktopts(opt, -1); /* XXX: discard all options */ 2229 free(opt, M_IP6OPT); 2230 return (error); 2231 } 2232 *pktopt = opt; 2233 return (0); 2234} 2235 2236/* 2237 * initialize ip6_pktopts. beware that there are non-zero default values in 2238 * the struct. 2239 */ 2240void 2241ip6_initpktopts(struct ip6_pktopts *opt) 2242{ 2243 2244 bzero(opt, sizeof(*opt)); 2245 opt->ip6po_hlim = -1; /* -1 means default hop limit */ 2246 opt->ip6po_tclass = -1; /* -1 means default traffic class */ 2247 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; 2248 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; 2249} 2250 2251static int 2252ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, 2253 struct ucred *cred, int uproto) 2254{ 2255 struct ip6_pktopts *opt; 2256 2257 if (*pktopt == NULL) { 2258 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT, 2259 M_WAITOK); 2260 ip6_initpktopts(*pktopt); 2261 } 2262 opt = *pktopt; 2263 2264 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto)); 2265} 2266 2267static int 2268ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) 2269{ 2270 void *optdata = NULL; 2271 int optdatalen = 0; 2272 struct ip6_ext *ip6e; 2273 int error = 0; 2274 struct in6_pktinfo null_pktinfo; 2275 int deftclass = 0, on; 2276 int defminmtu = IP6PO_MINMTU_MCASTONLY; 2277 int defpreftemp = IP6PO_TEMPADDR_SYSTEM; 2278 2279 switch (optname) { 2280 case IPV6_PKTINFO: 2281 optdata = (void *)&null_pktinfo; 2282 if (pktopt && pktopt->ip6po_pktinfo) { 2283 bcopy(pktopt->ip6po_pktinfo, &null_pktinfo, 2284 sizeof(null_pktinfo)); 2285 in6_clearscope(&null_pktinfo.ipi6_addr); 2286 } else { 2287 /* XXX: we don't have to do this every time... */ 2288 bzero(&null_pktinfo, sizeof(null_pktinfo)); 2289 } 2290 optdatalen = sizeof(struct in6_pktinfo); 2291 break; 2292 case IPV6_TCLASS: 2293 if (pktopt && pktopt->ip6po_tclass >= 0) 2294 optdata = (void *)&pktopt->ip6po_tclass; 2295 else 2296 optdata = (void *)&deftclass; 2297 optdatalen = sizeof(int); 2298 break; 2299 case IPV6_HOPOPTS: 2300 if (pktopt && pktopt->ip6po_hbh) { 2301 optdata = (void *)pktopt->ip6po_hbh; 2302 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; 2303 optdatalen = (ip6e->ip6e_len + 1) << 3; 2304 } 2305 break; 2306 case IPV6_RTHDR: 2307 if (pktopt && pktopt->ip6po_rthdr) { 2308 optdata = (void *)pktopt->ip6po_rthdr; 2309 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; 2310 optdatalen = (ip6e->ip6e_len + 1) << 3; 2311 } 2312 break; 2313 case IPV6_RTHDRDSTOPTS: 2314 if (pktopt && pktopt->ip6po_dest1) { 2315 optdata = (void *)pktopt->ip6po_dest1; 2316 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; 2317 optdatalen = (ip6e->ip6e_len + 1) << 3; 2318 } 2319 break; 2320 case IPV6_DSTOPTS: 2321 if (pktopt && pktopt->ip6po_dest2) { 2322 optdata = (void *)pktopt->ip6po_dest2; 2323 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; 2324 optdatalen = (ip6e->ip6e_len + 1) << 3; 2325 } 2326 break; 2327 case IPV6_NEXTHOP: 2328 if (pktopt && pktopt->ip6po_nexthop) { 2329 optdata = (void *)pktopt->ip6po_nexthop; 2330 optdatalen = pktopt->ip6po_nexthop->sa_len; 2331 } 2332 break; 2333 case IPV6_USE_MIN_MTU: 2334 if (pktopt) 2335 optdata = (void *)&pktopt->ip6po_minmtu; 2336 else 2337 optdata = (void *)&defminmtu; 2338 optdatalen = sizeof(int); 2339 break; 2340 case IPV6_DONTFRAG: 2341 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) 2342 on = 1; 2343 else 2344 on = 0; 2345 optdata = (void *)&on; 2346 optdatalen = sizeof(on); 2347 break; 2348 case IPV6_PREFER_TEMPADDR: 2349 if (pktopt) 2350 optdata = (void *)&pktopt->ip6po_prefer_tempaddr; 2351 else 2352 optdata = (void *)&defpreftemp; 2353 optdatalen = sizeof(int); 2354 break; 2355 default: /* should not happen */ 2356#ifdef DIAGNOSTIC 2357 panic("ip6_getpcbopt: unexpected option\n"); 2358#endif 2359 return (ENOPROTOOPT); 2360 } 2361 2362 error = sooptcopyout(sopt, optdata, optdatalen); 2363 2364 return (error); 2365} 2366 2367void 2368ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 2369{ 2370 if (pktopt == NULL) 2371 return; 2372 2373 if (optname == -1 || optname == IPV6_PKTINFO) { 2374 if (pktopt->ip6po_pktinfo) 2375 free(pktopt->ip6po_pktinfo, M_IP6OPT); 2376 pktopt->ip6po_pktinfo = NULL; 2377 } 2378 if (optname == -1 || optname == IPV6_HOPLIMIT) 2379 pktopt->ip6po_hlim = -1; 2380 if (optname == -1 || optname == IPV6_TCLASS) 2381 pktopt->ip6po_tclass = -1; 2382 if (optname == -1 || optname == IPV6_NEXTHOP) { 2383 if (pktopt->ip6po_nextroute.ro_rt) { 2384 RTFREE(pktopt->ip6po_nextroute.ro_rt); 2385 pktopt->ip6po_nextroute.ro_rt = NULL; 2386 } 2387 if (pktopt->ip6po_nexthop) 2388 free(pktopt->ip6po_nexthop, M_IP6OPT); 2389 pktopt->ip6po_nexthop = NULL; 2390 } 2391 if (optname == -1 || optname == IPV6_HOPOPTS) { 2392 if (pktopt->ip6po_hbh) 2393 free(pktopt->ip6po_hbh, M_IP6OPT); 2394 pktopt->ip6po_hbh = NULL; 2395 } 2396 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { 2397 if (pktopt->ip6po_dest1) 2398 free(pktopt->ip6po_dest1, M_IP6OPT); 2399 pktopt->ip6po_dest1 = NULL; 2400 } 2401 if (optname == -1 || optname == IPV6_RTHDR) { 2402 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) 2403 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); 2404 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; 2405 if (pktopt->ip6po_route.ro_rt) { 2406 RTFREE(pktopt->ip6po_route.ro_rt); 2407 pktopt->ip6po_route.ro_rt = NULL; 2408 } 2409 } 2410 if (optname == -1 || optname == IPV6_DSTOPTS) { 2411 if (pktopt->ip6po_dest2) 2412 free(pktopt->ip6po_dest2, M_IP6OPT); 2413 pktopt->ip6po_dest2 = NULL; 2414 } 2415} 2416 2417#define PKTOPT_EXTHDRCPY(type) \ 2418do {\ 2419 if (src->type) {\ 2420 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ 2421 dst->type = malloc(hlen, M_IP6OPT, canwait);\ 2422 if (dst->type == NULL && canwait == M_NOWAIT)\ 2423 goto bad;\ 2424 bcopy(src->type, dst->type, hlen);\ 2425 }\ 2426} while (/*CONSTCOND*/ 0) 2427 2428static int 2429copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 2430{ 2431 if (dst == NULL || src == NULL) { 2432 printf("ip6_clearpktopts: invalid argument\n"); 2433 return (EINVAL); 2434 } 2435 2436 dst->ip6po_hlim = src->ip6po_hlim; 2437 dst->ip6po_tclass = src->ip6po_tclass; 2438 dst->ip6po_flags = src->ip6po_flags; 2439 dst->ip6po_minmtu = src->ip6po_minmtu; 2440 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr; 2441 if (src->ip6po_pktinfo) { 2442 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo), 2443 M_IP6OPT, canwait); 2444 if (dst->ip6po_pktinfo == NULL) 2445 goto bad; 2446 *dst->ip6po_pktinfo = *src->ip6po_pktinfo; 2447 } 2448 if (src->ip6po_nexthop) { 2449 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len, 2450 M_IP6OPT, canwait); 2451 if (dst->ip6po_nexthop == NULL) 2452 goto bad; 2453 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, 2454 src->ip6po_nexthop->sa_len); 2455 } 2456 PKTOPT_EXTHDRCPY(ip6po_hbh); 2457 PKTOPT_EXTHDRCPY(ip6po_dest1); 2458 PKTOPT_EXTHDRCPY(ip6po_dest2); 2459 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ 2460 return (0); 2461 2462 bad: 2463 ip6_clearpktopts(dst, -1); 2464 return (ENOBUFS); 2465} 2466#undef PKTOPT_EXTHDRCPY 2467 2468struct ip6_pktopts * 2469ip6_copypktopts(struct ip6_pktopts *src, int canwait) 2470{ 2471 int error; 2472 struct ip6_pktopts *dst; 2473 2474 dst = malloc(sizeof(*dst), M_IP6OPT, canwait); 2475 if (dst == NULL) 2476 return (NULL); 2477 ip6_initpktopts(dst); 2478 2479 if ((error = copypktopts(dst, src, canwait)) != 0) { 2480 free(dst, M_IP6OPT); 2481 return (NULL); 2482 } 2483 2484 return (dst); 2485} 2486 2487void 2488ip6_freepcbopts(struct ip6_pktopts *pktopt) 2489{ 2490 if (pktopt == NULL) 2491 return; 2492 2493 ip6_clearpktopts(pktopt, -1); 2494 2495 free(pktopt, M_IP6OPT); 2496} 2497 2498/* 2499 * Set IPv6 outgoing packet options based on advanced API. 2500 */ 2501int 2502ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, 2503 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto) 2504{ 2505 struct cmsghdr *cm = NULL; 2506 2507 if (control == NULL || opt == NULL) 2508 return (EINVAL); 2509 2510 ip6_initpktopts(opt); 2511 if (stickyopt) { 2512 int error; 2513 2514 /* 2515 * If stickyopt is provided, make a local copy of the options 2516 * for this particular packet, then override them by ancillary 2517 * objects. 2518 * XXX: copypktopts() does not copy the cached route to a next 2519 * hop (if any). This is not very good in terms of efficiency, 2520 * but we can allow this since this option should be rarely 2521 * used. 2522 */ 2523 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) 2524 return (error); 2525 } 2526 2527 /* 2528 * XXX: Currently, we assume all the optional information is stored 2529 * in a single mbuf. 2530 */ 2531 if (control->m_next) 2532 return (EINVAL); 2533 2534 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), 2535 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 2536 int error; 2537 2538 if (control->m_len < CMSG_LEN(0)) 2539 return (EINVAL); 2540 2541 cm = mtod(control, struct cmsghdr *); 2542 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) 2543 return (EINVAL); 2544 if (cm->cmsg_level != IPPROTO_IPV6) 2545 continue; 2546 2547 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), 2548 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto); 2549 if (error) 2550 return (error); 2551 } 2552 2553 return (0); 2554} 2555 2556/* 2557 * Set a particular packet option, as a sticky option or an ancillary data 2558 * item. "len" can be 0 only when it's a sticky option. 2559 * We have 4 cases of combination of "sticky" and "cmsg": 2560 * "sticky=0, cmsg=0": impossible 2561 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data 2562 * "sticky=1, cmsg=0": RFC3542 socket option 2563 * "sticky=1, cmsg=1": RFC2292 socket option 2564 */ 2565static int 2566ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, 2567 struct ucred *cred, int sticky, int cmsg, int uproto) 2568{ 2569 int minmtupolicy, preftemp; 2570 int error; 2571 2572 if (!sticky && !cmsg) { 2573#ifdef DIAGNOSTIC 2574 printf("ip6_setpktopt: impossible case\n"); 2575#endif 2576 return (EINVAL); 2577 } 2578 2579 /* 2580 * IPV6_2292xxx is for backward compatibility to RFC2292, and should 2581 * not be specified in the context of RFC3542. Conversely, 2582 * RFC3542 types should not be specified in the context of RFC2292. 2583 */ 2584 if (!cmsg) { 2585 switch (optname) { 2586 case IPV6_2292PKTINFO: 2587 case IPV6_2292HOPLIMIT: 2588 case IPV6_2292NEXTHOP: 2589 case IPV6_2292HOPOPTS: 2590 case IPV6_2292DSTOPTS: 2591 case IPV6_2292RTHDR: 2592 case IPV6_2292PKTOPTIONS: 2593 return (ENOPROTOOPT); 2594 } 2595 } 2596 if (sticky && cmsg) { 2597 switch (optname) { 2598 case IPV6_PKTINFO: 2599 case IPV6_HOPLIMIT: 2600 case IPV6_NEXTHOP: 2601 case IPV6_HOPOPTS: 2602 case IPV6_DSTOPTS: 2603 case IPV6_RTHDRDSTOPTS: 2604 case IPV6_RTHDR: 2605 case IPV6_USE_MIN_MTU: 2606 case IPV6_DONTFRAG: 2607 case IPV6_TCLASS: 2608 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ 2609 return (ENOPROTOOPT); 2610 } 2611 } 2612 2613 switch (optname) { 2614 case IPV6_2292PKTINFO: 2615 case IPV6_PKTINFO: 2616 { 2617 struct ifnet *ifp = NULL; 2618 struct in6_pktinfo *pktinfo; 2619 2620 if (len != sizeof(struct in6_pktinfo)) 2621 return (EINVAL); 2622 2623 pktinfo = (struct in6_pktinfo *)buf; 2624 2625 /* 2626 * An application can clear any sticky IPV6_PKTINFO option by 2627 * doing a "regular" setsockopt with ipi6_addr being 2628 * in6addr_any and ipi6_ifindex being zero. 2629 * [RFC 3542, Section 6] 2630 */ 2631 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && 2632 pktinfo->ipi6_ifindex == 0 && 2633 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2634 ip6_clearpktopts(opt, optname); 2635 break; 2636 } 2637 2638 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && 2639 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2640 return (EINVAL); 2641 } 2642 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr)) 2643 return (EINVAL); 2644 /* validate the interface index if specified. */ 2645 if (pktinfo->ipi6_ifindex > V_if_index) 2646 return (ENXIO); 2647 if (pktinfo->ipi6_ifindex) { 2648 ifp = ifnet_byindex(pktinfo->ipi6_ifindex); 2649 if (ifp == NULL) 2650 return (ENXIO); 2651 } 2652 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL || 2653 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)) 2654 return (ENETDOWN); 2655 2656 if (ifp != NULL && 2657 !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { 2658 struct in6_ifaddr *ia; 2659 2660 in6_setscope(&pktinfo->ipi6_addr, ifp, NULL); 2661 ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr); 2662 if (ia == NULL) 2663 return (EADDRNOTAVAIL); 2664 ifa_free(&ia->ia_ifa); 2665 } 2666 /* 2667 * We store the address anyway, and let in6_selectsrc() 2668 * validate the specified address. This is because ipi6_addr 2669 * may not have enough information about its scope zone, and 2670 * we may need additional information (such as outgoing 2671 * interface or the scope zone of a destination address) to 2672 * disambiguate the scope. 2673 * XXX: the delay of the validation may confuse the 2674 * application when it is used as a sticky option. 2675 */ 2676 if (opt->ip6po_pktinfo == NULL) { 2677 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo), 2678 M_IP6OPT, M_NOWAIT); 2679 if (opt->ip6po_pktinfo == NULL) 2680 return (ENOBUFS); 2681 } 2682 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); 2683 break; 2684 } 2685 2686 case IPV6_2292HOPLIMIT: 2687 case IPV6_HOPLIMIT: 2688 { 2689 int *hlimp; 2690 2691 /* 2692 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT 2693 * to simplify the ordering among hoplimit options. 2694 */ 2695 if (optname == IPV6_HOPLIMIT && sticky) 2696 return (ENOPROTOOPT); 2697 2698 if (len != sizeof(int)) 2699 return (EINVAL); 2700 hlimp = (int *)buf; 2701 if (*hlimp < -1 || *hlimp > 255) 2702 return (EINVAL); 2703 2704 opt->ip6po_hlim = *hlimp; 2705 break; 2706 } 2707 2708 case IPV6_TCLASS: 2709 { 2710 int tclass; 2711 2712 if (len != sizeof(int)) 2713 return (EINVAL); 2714 tclass = *(int *)buf; 2715 if (tclass < -1 || tclass > 255) 2716 return (EINVAL); 2717 2718 opt->ip6po_tclass = tclass; 2719 break; 2720 } 2721 2722 case IPV6_2292NEXTHOP: 2723 case IPV6_NEXTHOP: 2724 if (cred != NULL) { 2725 error = priv_check_cred(cred, 2726 PRIV_NETINET_SETHDROPTS, 0); 2727 if (error) 2728 return (error); 2729 } 2730 2731 if (len == 0) { /* just remove the option */ 2732 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2733 break; 2734 } 2735 2736 /* check if cmsg_len is large enough for sa_len */ 2737 if (len < sizeof(struct sockaddr) || len < *buf) 2738 return (EINVAL); 2739 2740 switch (((struct sockaddr *)buf)->sa_family) { 2741 case AF_INET6: 2742 { 2743 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf; 2744 int error; 2745 2746 if (sa6->sin6_len != sizeof(struct sockaddr_in6)) 2747 return (EINVAL); 2748 2749 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 2750 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { 2751 return (EINVAL); 2752 } 2753 if ((error = sa6_embedscope(sa6, V_ip6_use_defzone)) 2754 != 0) { 2755 return (error); 2756 } 2757 break; 2758 } 2759 case AF_LINK: /* should eventually be supported */ 2760 default: 2761 return (EAFNOSUPPORT); 2762 } 2763 2764 /* turn off the previous option, then set the new option. */ 2765 ip6_clearpktopts(opt, IPV6_NEXTHOP); 2766 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT); 2767 if (opt->ip6po_nexthop == NULL) 2768 return (ENOBUFS); 2769 bcopy(buf, opt->ip6po_nexthop, *buf); 2770 break; 2771 2772 case IPV6_2292HOPOPTS: 2773 case IPV6_HOPOPTS: 2774 { 2775 struct ip6_hbh *hbh; 2776 int hbhlen; 2777 2778 /* 2779 * XXX: We don't allow a non-privileged user to set ANY HbH 2780 * options, since per-option restriction has too much 2781 * overhead. 2782 */ 2783 if (cred != NULL) { 2784 error = priv_check_cred(cred, 2785 PRIV_NETINET_SETHDROPTS, 0); 2786 if (error) 2787 return (error); 2788 } 2789 2790 if (len == 0) { 2791 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2792 break; /* just remove the option */ 2793 } 2794 2795 /* message length validation */ 2796 if (len < sizeof(struct ip6_hbh)) 2797 return (EINVAL); 2798 hbh = (struct ip6_hbh *)buf; 2799 hbhlen = (hbh->ip6h_len + 1) << 3; 2800 if (len != hbhlen) 2801 return (EINVAL); 2802 2803 /* turn off the previous option, then set the new option. */ 2804 ip6_clearpktopts(opt, IPV6_HOPOPTS); 2805 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT); 2806 if (opt->ip6po_hbh == NULL) 2807 return (ENOBUFS); 2808 bcopy(hbh, opt->ip6po_hbh, hbhlen); 2809 2810 break; 2811 } 2812 2813 case IPV6_2292DSTOPTS: 2814 case IPV6_DSTOPTS: 2815 case IPV6_RTHDRDSTOPTS: 2816 { 2817 struct ip6_dest *dest, **newdest = NULL; 2818 int destlen; 2819 2820 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */ 2821 error = priv_check_cred(cred, 2822 PRIV_NETINET_SETHDROPTS, 0); 2823 if (error) 2824 return (error); 2825 } 2826 2827 if (len == 0) { 2828 ip6_clearpktopts(opt, optname); 2829 break; /* just remove the option */ 2830 } 2831 2832 /* message length validation */ 2833 if (len < sizeof(struct ip6_dest)) 2834 return (EINVAL); 2835 dest = (struct ip6_dest *)buf; 2836 destlen = (dest->ip6d_len + 1) << 3; 2837 if (len != destlen) 2838 return (EINVAL); 2839 2840 /* 2841 * Determine the position that the destination options header 2842 * should be inserted; before or after the routing header. 2843 */ 2844 switch (optname) { 2845 case IPV6_2292DSTOPTS: 2846 /* 2847 * The old advacned API is ambiguous on this point. 2848 * Our approach is to determine the position based 2849 * according to the existence of a routing header. 2850 * Note, however, that this depends on the order of the 2851 * extension headers in the ancillary data; the 1st 2852 * part of the destination options header must appear 2853 * before the routing header in the ancillary data, 2854 * too. 2855 * RFC3542 solved the ambiguity by introducing 2856 * separate ancillary data or option types. 2857 */ 2858 if (opt->ip6po_rthdr == NULL) 2859 newdest = &opt->ip6po_dest1; 2860 else 2861 newdest = &opt->ip6po_dest2; 2862 break; 2863 case IPV6_RTHDRDSTOPTS: 2864 newdest = &opt->ip6po_dest1; 2865 break; 2866 case IPV6_DSTOPTS: 2867 newdest = &opt->ip6po_dest2; 2868 break; 2869 } 2870 2871 /* turn off the previous option, then set the new option. */ 2872 ip6_clearpktopts(opt, optname); 2873 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT); 2874 if (*newdest == NULL) 2875 return (ENOBUFS); 2876 bcopy(dest, *newdest, destlen); 2877 2878 break; 2879 } 2880 2881 case IPV6_2292RTHDR: 2882 case IPV6_RTHDR: 2883 { 2884 struct ip6_rthdr *rth; 2885 int rthlen; 2886 2887 if (len == 0) { 2888 ip6_clearpktopts(opt, IPV6_RTHDR); 2889 break; /* just remove the option */ 2890 } 2891 2892 /* message length validation */ 2893 if (len < sizeof(struct ip6_rthdr)) 2894 return (EINVAL); 2895 rth = (struct ip6_rthdr *)buf; 2896 rthlen = (rth->ip6r_len + 1) << 3; 2897 if (len != rthlen) 2898 return (EINVAL); 2899 2900 switch (rth->ip6r_type) { 2901 case IPV6_RTHDR_TYPE_0: 2902 if (rth->ip6r_len == 0) /* must contain one addr */ 2903 return (EINVAL); 2904 if (rth->ip6r_len % 2) /* length must be even */ 2905 return (EINVAL); 2906 if (rth->ip6r_len / 2 != rth->ip6r_segleft) 2907 return (EINVAL); 2908 break; 2909 default: 2910 return (EINVAL); /* not supported */ 2911 } 2912 2913 /* turn off the previous option */ 2914 ip6_clearpktopts(opt, IPV6_RTHDR); 2915 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT); 2916 if (opt->ip6po_rthdr == NULL) 2917 return (ENOBUFS); 2918 bcopy(rth, opt->ip6po_rthdr, rthlen); 2919 2920 break; 2921 } 2922 2923 case IPV6_USE_MIN_MTU: 2924 if (len != sizeof(int)) 2925 return (EINVAL); 2926 minmtupolicy = *(int *)buf; 2927 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && 2928 minmtupolicy != IP6PO_MINMTU_DISABLE && 2929 minmtupolicy != IP6PO_MINMTU_ALL) { 2930 return (EINVAL); 2931 } 2932 opt->ip6po_minmtu = minmtupolicy; 2933 break; 2934 2935 case IPV6_DONTFRAG: 2936 if (len != sizeof(int)) 2937 return (EINVAL); 2938 2939 if (uproto == IPPROTO_TCP || *(int *)buf == 0) { 2940 /* 2941 * we ignore this option for TCP sockets. 2942 * (RFC3542 leaves this case unspecified.) 2943 */ 2944 opt->ip6po_flags &= ~IP6PO_DONTFRAG; 2945 } else 2946 opt->ip6po_flags |= IP6PO_DONTFRAG; 2947 break; 2948 2949 case IPV6_PREFER_TEMPADDR: 2950 if (len != sizeof(int)) 2951 return (EINVAL); 2952 preftemp = *(int *)buf; 2953 if (preftemp != IP6PO_TEMPADDR_SYSTEM && 2954 preftemp != IP6PO_TEMPADDR_NOTPREFER && 2955 preftemp != IP6PO_TEMPADDR_PREFER) { 2956 return (EINVAL); 2957 } 2958 opt->ip6po_prefer_tempaddr = preftemp; 2959 break; 2960 2961 default: 2962 return (ENOPROTOOPT); 2963 } /* end of switch */ 2964 2965 return (0); 2966} 2967 2968/* 2969 * Routine called from ip6_output() to loop back a copy of an IP6 multicast 2970 * packet to the input queue of a specified interface. Note that this 2971 * calls the output routine of the loopback "driver", but with an interface 2972 * pointer that might NOT be &loif -- easier than replicating that code here. 2973 */ 2974void 2975ip6_mloopback(struct ifnet *ifp, struct mbuf *m) 2976{ 2977 struct mbuf *copym; 2978 struct ip6_hdr *ip6; 2979 2980 copym = m_copy(m, 0, M_COPYALL); 2981 if (copym == NULL) 2982 return; 2983 2984 /* 2985 * Make sure to deep-copy IPv6 header portion in case the data 2986 * is in an mbuf cluster, so that we can safely override the IPv6 2987 * header portion later. 2988 */ 2989 if (!M_WRITABLE(copym) || 2990 copym->m_len < sizeof(struct ip6_hdr)) { 2991 copym = m_pullup(copym, sizeof(struct ip6_hdr)); 2992 if (copym == NULL) 2993 return; 2994 } 2995 ip6 = mtod(copym, struct ip6_hdr *); 2996 /* 2997 * clear embedded scope identifiers if necessary. 2998 * in6_clearscope will touch the addresses only when necessary. 2999 */ 3000 in6_clearscope(&ip6->ip6_src); 3001 in6_clearscope(&ip6->ip6_dst); 3002 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 3003 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 | 3004 CSUM_PSEUDO_HDR; 3005 copym->m_pkthdr.csum_data = 0xffff; 3006 } 3007 if_simloop(ifp, copym, AF_INET6, 0); 3008} 3009 3010/* 3011 * Chop IPv6 header off from the payload. 3012 */ 3013static int 3014ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) 3015{ 3016 struct mbuf *mh; 3017 struct ip6_hdr *ip6; 3018 3019 ip6 = mtod(m, struct ip6_hdr *); 3020 if (m->m_len > sizeof(*ip6)) { 3021 mh = m_gethdr(M_NOWAIT, MT_DATA); 3022 if (mh == NULL) { 3023 m_freem(m); 3024 return ENOBUFS; 3025 } 3026 m_move_pkthdr(mh, m); 3027 M_ALIGN(mh, sizeof(*ip6)); 3028 m->m_len -= sizeof(*ip6); 3029 m->m_data += sizeof(*ip6); 3030 mh->m_next = m; 3031 m = mh; 3032 m->m_len = sizeof(*ip6); 3033 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); 3034 } 3035 exthdrs->ip6e_ip6 = m; 3036 return 0; 3037} 3038 3039/* 3040 * Compute IPv6 extension header length. 3041 */ 3042int 3043ip6_optlen(struct inpcb *in6p) 3044{ 3045 int len; 3046 3047 if (!in6p->in6p_outputopts) 3048 return 0; 3049 3050 len = 0; 3051#define elen(x) \ 3052 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) 3053 3054 len += elen(in6p->in6p_outputopts->ip6po_hbh); 3055 if (in6p->in6p_outputopts->ip6po_rthdr) 3056 /* dest1 is valid with rthdr only */ 3057 len += elen(in6p->in6p_outputopts->ip6po_dest1); 3058 len += elen(in6p->in6p_outputopts->ip6po_rthdr); 3059 len += elen(in6p->in6p_outputopts->ip6po_dest2); 3060 return len; 3061#undef elen 3062} 3063